The University of Montreal last year trumpeted a new partnership with a leading genealogical organization that would index some 12-million records from old French-Canadian civil records. Cooperation between the French-language university and ancestry.ca was called a “significant accomplishment” by U of M senior researcher Bertrand Desjardins in a posting on the university’s website, given that it would offer researchers access to records containing 37-million French-Canadian names and 3.6-million images from the comprehensive Drouin collection, going back to 1621.
But frustration has set in among family-tree researchers thanks to ancestry.ca’s decision to outsource the project to China. The Chinese workers, it seems, have made numerous errors in transcribing the baptism, marriage and burial records.
The university posting acknowledged the errors and indicated that The Generations Network, the American holding company of Ancestry.ca, was contributing $200,000 to the university to hire U of M students to wade through the files to make corrections. The process would take five years. The outsourcing was done, the site admits, to save money.
Desjardins appeared to laugh off the gaffes in the website posting: “French-Canadian names were transcribed into databases by the Chinese who sometimes made mistakes. For instance, ‘Charbonneau’ was spelled ‘Chorbanneau,’ explained Desjar-dins amusingly.”
But researchers at iforum.umontreal.ca were not as amused as Desjardins that the records as posted on the internet were so flawed. Why upload them, they argued, and charge fees, if they are not accurate or checked?
Any French speaker undertaking the challenge of poring through old Quebec records will be familiar with the inconsistencies in recording facts in the first place, with spelling and names altered from generation to generation, and frequently the word dit (“called”) changing a surname. New variations introduced through errors in the 21st century are not welcomed.
Researchers already have a gold-standard but smaller database, known as the PRDH, to compare the outsourced content to in terms of quality. The Chinese transcribers seemed to have trouble with the nomenclature, the culture, the handwriting on the records and the language itself. “Un enfant” was transcribed as a first name whereas it means “a child” in English. “Drouin” became “Browne.” One record is for a “Montreal Bilodeau,” clearly a mistake.
One forum contributor identifying himself as Andrew Cunningham commented in April, “The Ancestry version ... is so full of errors and omissions that one has to take very special care to double-check the original books and other sources for accuracy … Do not depend on this index – it is not accurate!”
Desjardins defended the outsourcing via email:
“I realize that the indexing work done by people who do not have any idea of what they are keying is far from ideal. But let's look at the numbers here. The Drouin collection contains some 12-million records. They are hand-written, with the different variables of primary interest – date, age, names of the subjects and of their parents and spouse, place of origin – not being written in fixed fields which would facilitate their identification. Keying such a number of items is a mammoth task. Everyone would prefer to have a PRDH quality index, but who would be ready to pay for the costs involved?
“It is thus a simple question of economics. The Generations Network, parent company of the Ancestry sites, chose to limit costs by outsourcing the keying and limiting it to the marginal indications of the records, because it made business sense ... within a year, a new tool was made available to researchers. Far from ideal? Definitely. Useful? Definitely!”